Understanding sweep jobs

Sweep jobs in AMLS enable a data scientist to define the hyperparameters to explore in a single job. During the job, this will automate the task of searching for the hyperparameters that will provide a model with the best results for the primary metric-creating trials. In a run of a job, multiple trials are created and evaluated for the hyperparameters that are defined within the search space based on the sampling method selected. By defining the search space, we can create a single run of a job for testing multiple hypotheses at a single time rather than re-writing code and re-running jobs, reducing the time spent exploring the search space.

To leverage the hyperparameters in your job, your code needs to be updated to leverage these new parameters by passing them into your code through the Python ArgumentParser shown as follows:

Figure 4.6 – Passing a parameter list into the job

Now that the arguments have been passed...